VGleaks: Orbis Unveiled! [Updated]

Those extra CU's might also be used for rendering so they need to have the same feature sets, also mapping vectors and matrices to textures buys you quite a bit of compute performance, seen it first hand years ago with CUDA kernels.

It does make sense if SCE asked for the HW scheduler in those extra CU's and the branch predictor, amongst other bits, to be optimized more for compute tasks than graphics processing so to offer more help to the CPU.

Sorry, not much of a tech head but I was wondering if these 4 CU's, which are more on the CPU's side, could be used for tessellation?

My limited understanding is normally the basic mesh goes from CPU to GPU, then the GPU handles tessellation. If these CU's are used for compute functions, it would make sense (to me at least) that these CU's do the heavy lifting all on the CPU's side of the APU, without clogging up the works between CPU & GPU

It would also make a bit of sense when they used the term "hardware balanced"
 
Do we have any info on tesselation performance on Orbis? That and realtime lighting are going to be absolutely key for next gen.
 
Do we have any info on tesselation performance on Orbis? That and realtime lighting are going to be absolutely key for next gen.

GCN should have standard tesselation support in the triangle setup unit - I don't think it varies by the number of CUs, so it should have similar triangle setup and tesselation performance as a 7950 I think, its a fixed function unit.

might be possible to augment that with the 'spare' CUs if it isn't enough or you want to do something special
 
GCN should have standard tesselation support in the triangle setup unit - I don't think it varies by the number of CUs, so it should have similar triangle setup and tesselation performance as a 7950 I think, its a fixed function unit.

might be possible to augment that with the 'spare' CUs if it isn't enough or you want to do something special
Makes sense. I assume the tesselation performance of a 7950 is pretty good? I guess it hasn't been used quite enough yet to really tell.
 
Makes sense. I assume the tesselation performance of a 7950 is pretty good? I guess it hasn't bren used quite enough yet to really tell.

yes and yes. Its one area I'm interested in seeing be exploited - PC games seem to be mostly tinkering round the edges.

79xx series is supposed to be 'up to 4x' tesselation performance compared to 6xxx series.

there are some good white papers and developer docs online, they're fairly readable.
 
XDR was about the same speed as GDDR5. But memory speed on a cpu has no really any performance gain since ddr2.
quake4.PNG

This whole post is all sorts of bad.

Try putting GDDR5 into a PS3 or 360 in 2006, see how much that costs and how much heat issues they'd get (more than they already have). XDR runs significantly cooler.

Also, Cell isn't a "CPU". It is very bandwidth limited. XDR was as cheap as they can get with the low power requirements and high bandwidth speeds.

I don't know how you can ignore such things, but ok.
 
Pretty big difference honestly.

It's looking more and more like Sony was right, and bandwidth is the key issue.

This is trivial to an extent though.

The GPU portion is obviously going to show massive increases in performance due to how hamstrung it is bandwidth wise (128bit bus means as much MHz as it can get is ideal) relative to its shading power.

If you look at the 1080p results too, the difference is minimal versus 720p.

For our purely academic purposes, we could do with them experimenting with various RAM pools too.
 
Consoles may be a lot more efficient and have extra hardware, but I still don't see how a 1.2 TF GPU should be outperforming a 4+ TFLOPs GPU.

In any case, I hope some of the extra power is put into image quality (shameless copy from Crysis 3 beta thread)

PS3/PC high

If the unthinkable does happen... Reserve your meltdown for E3.
 
Yeah if you type any file other than html, and the server thought there is a file which it is not there and will redirect to index.html. It seen it only happen same path as index.html. Sub-url will redirect to 404.
 
I just want to comment on this - IMHO, everything that is shitty about the PS3 architecture you can pretty much blame on the RSX.

Shitty compared to what though? I think that's what is missing from that equation. Shitty compared to the other GPU options at the time? Not at all. Shitty compared to Xenos? That's an exaggeration but that has more to do with AMD providing a better than expected GPU than Nvidia providing a lackluster GPU.

Consoles may be a lot more efficient and have extra hardware, but I still don't see how a 1.2 TF GPU should be outperforming a 4+ TFLOPs GPU.

In any case, I hope some of the extra power is put into image quality (shameless copy from Crysis 3 beta thread)

PS3/PC high

I really hope we get a Crysis 3 re-release on next gen consoles. If I knew it was a launch title for Orbis, I'd skip the current gen versions and just wait for that.
 
Memory speed effect on APU A10 5800k test:

http://www.legitreviews.com/article/2106/1/

A modified or similar version of this APU is reportedly in dev kits. Memory speed does have a noticible effect on performance.



Looks like memory speed is even more important now than 3-4 years ago. It preatty much sums what i was talking about. With next gen games amount of things on screen will need fast ram. Without it your hardware will be starved.

30% difference is big deal. With closed system to metal coding it will be even more important.

4GB GDDR5 was 10/10 shot from Sony
 
Consoles may be a lot more efficient and have extra hardware, but I still don't see how a 1.2 TF GPU should be outperforming a 4+ TFLOPs GPU.
It won't, but PC GPUs are severely underutilized. You can max out Crysis 2 on something like 800GFLOps GPU in terms of effects etc., and it's still probably the prettiest PC game. 1.2TFLOps should be easily capable of providing imagery pretty enough for the general audience, obviously gaming enthusiasts will stick to PCs but we've known it'd be this way ever since MS and Sony wasted billions while Nintendo was outselling them handily in 2006-2009.
 
Good news: sony joins the HSA Foundation:
http://www.fudzilla.com/home/item/30278-hsa-foundation-gets-few-more-members

The HSA (Heterogeneous System Architecture) Foundation is a not-for-profit consortium for SoC IP vendors, OEMs, academia, SoC vendors, OSVs and ISVs whose goal is to make it easy to program for parallel computing. HSA members are building a heterogeneous compute ecosystem, rooted in industry standards, for combining scalar processing on the CPU with parallel processing on the GPU while enabling high bandwidth access to memory and high application performance at low power consumption. HSA defines interfaces for parallel computation utilizing CPU, GPU and other programmable and fixed function devices, and support for a diverse set of high-level programming languages, thereby creating the next foundation in general purpose computing.
 
The worst part of these new generations is that eveyone thinks they engineered these systems. No one has a clue what's actually happening behind the scenes except the people working on these machines. You idiots think MS is going to intentionally gimp their system? Reserve your judgment until official specs are released and games are shown.
 
Googling, from Sotfopedia:

Jaguar now comes with most – if not all – of Bobcat’s abilities, but it also adds complete architectural computational support for SSE4.1, SSE4.2, AES, CLMUL, MOVBE, AVX, XSAVE, XSAVEOPT, FC16, and BMI instructions.

Since most SIMD instructions are 128-bit wide – if not wider, AMD decided to double the width of its FPU pipelines from 64-bit to 128-bit.

SSE instructions required two passes on the Bobcat, but they now can be handled in one single pass, and this is no small feat.

Ironically for AMD, Jaguar comes with support for AVX and these would be best fit by 256-bit wide FPU pipelines,but this is not the main concern for a mobile-oriented architecture.

I was trying to start a discussion about this back on the 18th, no one was interested.

http://www.neogaf.com/forum/showpost.php?p=46597467&postcount=7419
 
Top Bottom